blob: 48386bff8b4e4276d35775a5e995444860bb39a2 [file] [log] [blame]
Joe Stringer7f8a4362015-08-26 11:31:48 -07001/*
2 * Copyright (c) 2015 Nicira, Inc.
3 *
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of version 2 of the GNU General Public
6 * License as published by the Free Software Foundation.
7 *
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
12 */
13
14#include <linux/module.h>
15#include <linux/openvswitch.h>
Jarno Rajahalme05752522016-03-10 10:54:23 -080016#include <linux/tcp.h>
17#include <linux/udp.h>
18#include <linux/sctp.h>
Joe Stringer7f8a4362015-08-26 11:31:48 -070019#include <net/ip.h>
20#include <net/netfilter/nf_conntrack_core.h>
Joe Stringercae3a262015-08-26 11:31:53 -070021#include <net/netfilter/nf_conntrack_helper.h>
Joe Stringerc2ac6672015-08-26 11:31:52 -070022#include <net/netfilter/nf_conntrack_labels.h>
Jarno Rajahalme05752522016-03-10 10:54:23 -080023#include <net/netfilter/nf_conntrack_seqadj.h>
Joe Stringer7f8a4362015-08-26 11:31:48 -070024#include <net/netfilter/nf_conntrack_zones.h>
25#include <net/netfilter/ipv6/nf_defrag_ipv6.h>
26
Jarno Rajahalme05752522016-03-10 10:54:23 -080027#ifdef CONFIG_NF_NAT_NEEDED
28#include <linux/netfilter/nf_nat.h>
29#include <net/netfilter/nf_nat_core.h>
30#include <net/netfilter/nf_nat_l3proto.h>
31#endif
32
Joe Stringer7f8a4362015-08-26 11:31:48 -070033#include "datapath.h"
34#include "conntrack.h"
35#include "flow.h"
36#include "flow_netlink.h"
37
38struct ovs_ct_len_tbl {
Jarno Rajahalme05752522016-03-10 10:54:23 -080039 int maxlen;
40 int minlen;
Joe Stringer7f8a4362015-08-26 11:31:48 -070041};
42
Joe Stringer182e3042015-08-26 11:31:49 -070043/* Metadata mark for masked write to conntrack mark */
44struct md_mark {
45 u32 value;
46 u32 mask;
47};
48
Joe Stringerc2ac6672015-08-26 11:31:52 -070049/* Metadata label for masked write to conntrack label. */
Joe Stringer33db4122015-10-01 15:00:37 -070050struct md_labels {
51 struct ovs_key_ct_labels value;
52 struct ovs_key_ct_labels mask;
Joe Stringerc2ac6672015-08-26 11:31:52 -070053};
54
Jarno Rajahalme05752522016-03-10 10:54:23 -080055enum ovs_ct_nat {
56 OVS_CT_NAT = 1 << 0, /* NAT for committed connections only. */
57 OVS_CT_SRC_NAT = 1 << 1, /* Source NAT for NEW connections. */
58 OVS_CT_DST_NAT = 1 << 2, /* Destination NAT for NEW connections. */
59};
60
Joe Stringer7f8a4362015-08-26 11:31:48 -070061/* Conntrack action context for execution. */
62struct ovs_conntrack_info {
Joe Stringercae3a262015-08-26 11:31:53 -070063 struct nf_conntrack_helper *helper;
Joe Stringer7f8a4362015-08-26 11:31:48 -070064 struct nf_conntrack_zone zone;
65 struct nf_conn *ct;
Joe Stringerab38a7b2015-10-06 11:00:01 -070066 u8 commit : 1;
Jarno Rajahalme05752522016-03-10 10:54:23 -080067 u8 nat : 3; /* enum ovs_ct_nat */
Joe Stringer7f8a4362015-08-26 11:31:48 -070068 u16 family;
Joe Stringer182e3042015-08-26 11:31:49 -070069 struct md_mark mark;
Joe Stringer33db4122015-10-01 15:00:37 -070070 struct md_labels labels;
Jarno Rajahalme05752522016-03-10 10:54:23 -080071#ifdef CONFIG_NF_NAT_NEEDED
72 struct nf_nat_range range; /* Only present for SRC NAT and DST NAT. */
73#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -070074};
75
Joe Stringer2f3ab9f2015-12-09 14:07:39 -080076static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info);
77
Joe Stringer7f8a4362015-08-26 11:31:48 -070078static u16 key_to_nfproto(const struct sw_flow_key *key)
79{
80 switch (ntohs(key->eth.type)) {
81 case ETH_P_IP:
82 return NFPROTO_IPV4;
83 case ETH_P_IPV6:
84 return NFPROTO_IPV6;
85 default:
86 return NFPROTO_UNSPEC;
87 }
88}
89
90/* Map SKB connection state into the values used by flow definition. */
91static u8 ovs_ct_get_state(enum ip_conntrack_info ctinfo)
92{
93 u8 ct_state = OVS_CS_F_TRACKED;
94
95 switch (ctinfo) {
96 case IP_CT_ESTABLISHED_REPLY:
97 case IP_CT_RELATED_REPLY:
Joe Stringer7f8a4362015-08-26 11:31:48 -070098 ct_state |= OVS_CS_F_REPLY_DIR;
99 break;
100 default:
101 break;
102 }
103
104 switch (ctinfo) {
105 case IP_CT_ESTABLISHED:
106 case IP_CT_ESTABLISHED_REPLY:
107 ct_state |= OVS_CS_F_ESTABLISHED;
108 break;
109 case IP_CT_RELATED:
110 case IP_CT_RELATED_REPLY:
111 ct_state |= OVS_CS_F_RELATED;
112 break;
113 case IP_CT_NEW:
Joe Stringer7f8a4362015-08-26 11:31:48 -0700114 ct_state |= OVS_CS_F_NEW;
115 break;
116 default:
117 break;
118 }
119
120 return ct_state;
121}
122
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700123static u32 ovs_ct_get_mark(const struct nf_conn *ct)
124{
125#if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
126 return ct ? ct->mark : 0;
127#else
128 return 0;
129#endif
130}
131
Joe Stringer33db4122015-10-01 15:00:37 -0700132static void ovs_ct_get_labels(const struct nf_conn *ct,
133 struct ovs_key_ct_labels *labels)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700134{
135 struct nf_conn_labels *cl = ct ? nf_ct_labels_find(ct) : NULL;
136
137 if (cl) {
Florian Westphal23014012016-07-21 12:51:16 +0200138 size_t len = sizeof(cl->bits);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700139
Joe Stringer33db4122015-10-01 15:00:37 -0700140 if (len > OVS_CT_LABELS_LEN)
141 len = OVS_CT_LABELS_LEN;
142 else if (len < OVS_CT_LABELS_LEN)
143 memset(labels, 0, OVS_CT_LABELS_LEN);
144 memcpy(labels, cl->bits, len);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700145 } else {
Joe Stringer33db4122015-10-01 15:00:37 -0700146 memset(labels, 0, OVS_CT_LABELS_LEN);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700147 }
148}
149
Joe Stringer7f8a4362015-08-26 11:31:48 -0700150static void __ovs_ct_update_key(struct sw_flow_key *key, u8 state,
Joe Stringer182e3042015-08-26 11:31:49 -0700151 const struct nf_conntrack_zone *zone,
152 const struct nf_conn *ct)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700153{
154 key->ct.state = state;
155 key->ct.zone = zone->id;
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700156 key->ct.mark = ovs_ct_get_mark(ct);
Joe Stringer33db4122015-10-01 15:00:37 -0700157 ovs_ct_get_labels(ct, &key->ct.labels);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700158}
159
Jarno Rajahalme05752522016-03-10 10:54:23 -0800160/* Update 'key' based on skb->nfct. If 'post_ct' is true, then OVS has
161 * previously sent the packet to conntrack via the ct action. If
162 * 'keep_nat_flags' is true, the existing NAT flags retained, else they are
163 * initialized from the connection status.
Joe Stringer7f8a4362015-08-26 11:31:48 -0700164 */
165static void ovs_ct_update_key(const struct sk_buff *skb,
Joe Stringerd1109862015-12-09 14:07:40 -0800166 const struct ovs_conntrack_info *info,
Jarno Rajahalme05752522016-03-10 10:54:23 -0800167 struct sw_flow_key *key, bool post_ct,
168 bool keep_nat_flags)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700169{
170 const struct nf_conntrack_zone *zone = &nf_ct_zone_dflt;
171 enum ip_conntrack_info ctinfo;
172 struct nf_conn *ct;
173 u8 state = 0;
174
175 ct = nf_ct_get(skb, &ctinfo);
176 if (ct) {
177 state = ovs_ct_get_state(ctinfo);
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800178 /* All unconfirmed entries are NEW connections. */
Joe Stringer4f0909e2015-10-19 19:18:59 -0700179 if (!nf_ct_is_confirmed(ct))
180 state |= OVS_CS_F_NEW;
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800181 /* OVS persists the related flag for the duration of the
182 * connection.
183 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700184 if (ct->master)
185 state |= OVS_CS_F_RELATED;
Jarno Rajahalme05752522016-03-10 10:54:23 -0800186 if (keep_nat_flags) {
187 state |= key->ct.state & OVS_CS_F_NAT_MASK;
188 } else {
189 if (ct->status & IPS_SRC_NAT)
190 state |= OVS_CS_F_SRC_NAT;
191 if (ct->status & IPS_DST_NAT)
192 state |= OVS_CS_F_DST_NAT;
193 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700194 zone = nf_ct_zone(ct);
195 } else if (post_ct) {
196 state = OVS_CS_F_TRACKED | OVS_CS_F_INVALID;
Joe Stringerd1109862015-12-09 14:07:40 -0800197 if (info)
198 zone = &info->zone;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700199 }
Joe Stringer182e3042015-08-26 11:31:49 -0700200 __ovs_ct_update_key(key, state, zone, ct);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700201}
202
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800203/* This is called to initialize CT key fields possibly coming in from the local
204 * stack.
205 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700206void ovs_ct_fill_key(const struct sk_buff *skb, struct sw_flow_key *key)
207{
Jarno Rajahalme05752522016-03-10 10:54:23 -0800208 ovs_ct_update_key(skb, NULL, key, false, false);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700209}
210
211int ovs_ct_put_key(const struct sw_flow_key *key, struct sk_buff *skb)
212{
Joe Stringerfbccce52015-10-06 11:00:00 -0700213 if (nla_put_u32(skb, OVS_KEY_ATTR_CT_STATE, key->ct.state))
Joe Stringer7f8a4362015-08-26 11:31:48 -0700214 return -EMSGSIZE;
215
216 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
217 nla_put_u16(skb, OVS_KEY_ATTR_CT_ZONE, key->ct.zone))
218 return -EMSGSIZE;
219
Joe Stringer182e3042015-08-26 11:31:49 -0700220 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
221 nla_put_u32(skb, OVS_KEY_ATTR_CT_MARK, key->ct.mark))
222 return -EMSGSIZE;
223
Valentin Rothberg9723e6a2015-08-28 10:39:56 +0200224 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringer33db4122015-10-01 15:00:37 -0700225 nla_put(skb, OVS_KEY_ATTR_CT_LABELS, sizeof(key->ct.labels),
226 &key->ct.labels))
Joe Stringerc2ac6672015-08-26 11:31:52 -0700227 return -EMSGSIZE;
228
Joe Stringer182e3042015-08-26 11:31:49 -0700229 return 0;
230}
231
232static int ovs_ct_set_mark(struct sk_buff *skb, struct sw_flow_key *key,
233 u32 ct_mark, u32 mask)
234{
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700235#if IS_ENABLED(CONFIG_NF_CONNTRACK_MARK)
Joe Stringer182e3042015-08-26 11:31:49 -0700236 enum ip_conntrack_info ctinfo;
237 struct nf_conn *ct;
238 u32 new_mark;
239
Joe Stringer182e3042015-08-26 11:31:49 -0700240 /* The connection could be invalid, in which case set_mark is no-op. */
241 ct = nf_ct_get(skb, &ctinfo);
242 if (!ct)
243 return 0;
244
245 new_mark = ct_mark | (ct->mark & ~(mask));
246 if (ct->mark != new_mark) {
247 ct->mark = new_mark;
248 nf_conntrack_event_cache(IPCT_MARK, ct);
249 key->ct.mark = new_mark;
250 }
251
Joe Stringer7f8a4362015-08-26 11:31:48 -0700252 return 0;
Joe Stringer0d5cdef2015-08-28 19:22:11 -0700253#else
254 return -ENOTSUPP;
255#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -0700256}
257
Joe Stringer33db4122015-10-01 15:00:37 -0700258static int ovs_ct_set_labels(struct sk_buff *skb, struct sw_flow_key *key,
259 const struct ovs_key_ct_labels *labels,
260 const struct ovs_key_ct_labels *mask)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700261{
262 enum ip_conntrack_info ctinfo;
263 struct nf_conn_labels *cl;
264 struct nf_conn *ct;
265 int err;
266
Joe Stringerc2ac6672015-08-26 11:31:52 -0700267 /* The connection could be invalid, in which case set_label is no-op.*/
268 ct = nf_ct_get(skb, &ctinfo);
269 if (!ct)
270 return 0;
271
272 cl = nf_ct_labels_find(ct);
273 if (!cl) {
274 nf_ct_labels_ext_add(ct);
275 cl = nf_ct_labels_find(ct);
276 }
Florian Westphal23014012016-07-21 12:51:16 +0200277 if (!cl || sizeof(cl->bits) < OVS_CT_LABELS_LEN)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700278 return -ENOSPC;
279
Joe Stringer33db4122015-10-01 15:00:37 -0700280 err = nf_connlabels_replace(ct, (u32 *)labels, (u32 *)mask,
281 OVS_CT_LABELS_LEN / sizeof(u32));
Joe Stringerc2ac6672015-08-26 11:31:52 -0700282 if (err)
283 return err;
284
Joe Stringer33db4122015-10-01 15:00:37 -0700285 ovs_ct_get_labels(ct, &key->ct.labels);
Joe Stringerc2ac6672015-08-26 11:31:52 -0700286 return 0;
287}
288
Joe Stringercae3a262015-08-26 11:31:53 -0700289/* 'skb' should already be pulled to nh_ofs. */
290static int ovs_ct_helper(struct sk_buff *skb, u16 proto)
291{
292 const struct nf_conntrack_helper *helper;
293 const struct nf_conn_help *help;
294 enum ip_conntrack_info ctinfo;
295 unsigned int protoff;
296 struct nf_conn *ct;
Jarno Rajahalme05752522016-03-10 10:54:23 -0800297 int err;
Joe Stringercae3a262015-08-26 11:31:53 -0700298
299 ct = nf_ct_get(skb, &ctinfo);
300 if (!ct || ctinfo == IP_CT_RELATED_REPLY)
301 return NF_ACCEPT;
302
303 help = nfct_help(ct);
304 if (!help)
305 return NF_ACCEPT;
306
307 helper = rcu_dereference(help->helper);
308 if (!helper)
309 return NF_ACCEPT;
310
311 switch (proto) {
312 case NFPROTO_IPV4:
313 protoff = ip_hdrlen(skb);
314 break;
315 case NFPROTO_IPV6: {
316 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
317 __be16 frag_off;
Joe Stringercc570602015-09-14 11:14:50 -0700318 int ofs;
Joe Stringercae3a262015-08-26 11:31:53 -0700319
Joe Stringercc570602015-09-14 11:14:50 -0700320 ofs = ipv6_skip_exthdr(skb, sizeof(struct ipv6hdr), &nexthdr,
321 &frag_off);
322 if (ofs < 0 || (frag_off & htons(~0x7)) != 0) {
Joe Stringercae3a262015-08-26 11:31:53 -0700323 pr_debug("proto header not found\n");
324 return NF_ACCEPT;
325 }
Joe Stringercc570602015-09-14 11:14:50 -0700326 protoff = ofs;
Joe Stringercae3a262015-08-26 11:31:53 -0700327 break;
328 }
329 default:
330 WARN_ONCE(1, "helper invoked on non-IP family!");
331 return NF_DROP;
332 }
333
Jarno Rajahalme05752522016-03-10 10:54:23 -0800334 err = helper->help(skb, protoff, ct, ctinfo);
335 if (err != NF_ACCEPT)
336 return err;
337
338 /* Adjust seqs after helper. This is needed due to some helpers (e.g.,
339 * FTP with NAT) adusting the TCP payload size when mangling IP
340 * addresses and/or port numbers in the text-based control connection.
341 */
342 if (test_bit(IPS_SEQ_ADJUST_BIT, &ct->status) &&
343 !nf_ct_seq_adjust(skb, ct, ctinfo, protoff))
344 return NF_DROP;
345 return NF_ACCEPT;
Joe Stringercae3a262015-08-26 11:31:53 -0700346}
347
Joe Stringer74c16612015-10-25 20:21:48 -0700348/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
349 * value if 'skb' is freed.
350 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700351static int handle_fragments(struct net *net, struct sw_flow_key *key,
352 u16 zone, struct sk_buff *skb)
353{
354 struct ovs_skb_cb ovs_cb = *OVS_CB(skb);
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100355 int err;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700356
357 if (key->eth.type == htons(ETH_P_IP)) {
358 enum ip_defrag_users user = IP_DEFRAG_CONNTRACK_IN + zone;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700359
360 memset(IPCB(skb), 0, sizeof(struct inet_skb_parm));
Eric W. Biederman19bcf9f2015-10-09 13:44:54 -0500361 err = ip_defrag(net, skb, user);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700362 if (err)
363 return err;
364
365 ovs_cb.mru = IPCB(skb)->frag_max_size;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700366#if IS_ENABLED(CONFIG_NF_DEFRAG_IPV6)
Joe Stringer74c16612015-10-25 20:21:48 -0700367 } else if (key->eth.type == htons(ETH_P_IPV6)) {
Joe Stringer7f8a4362015-08-26 11:31:48 -0700368 enum ip6_defrag_users user = IP6_DEFRAG_CONNTRACK_IN + zone;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700369
370 memset(IP6CB(skb), 0, sizeof(struct inet6_skb_parm));
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100371 err = nf_ct_frag6_gather(net, skb, user);
Daniele Di Proiettof92a80a2016-11-28 15:43:53 -0800372 if (err) {
373 if (err != -EINPROGRESS)
374 kfree_skb(skb);
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100375 return err;
Daniele Di Proiettof92a80a2016-11-28 15:43:53 -0800376 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700377
Florian Westphaldaaa7d62015-11-18 23:32:40 +0100378 key->ip.proto = ipv6_hdr(skb)->nexthdr;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700379 ovs_cb.mru = IP6CB(skb)->frag_max_size;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700380#endif
381 } else {
Joe Stringer74c16612015-10-25 20:21:48 -0700382 kfree_skb(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700383 return -EPFNOSUPPORT;
384 }
385
386 key->ip.frag = OVS_FRAG_TYPE_NONE;
387 skb_clear_hash(skb);
388 skb->ignore_df = 1;
389 *OVS_CB(skb) = ovs_cb;
390
391 return 0;
392}
393
394static struct nf_conntrack_expect *
395ovs_ct_expect_find(struct net *net, const struct nf_conntrack_zone *zone,
396 u16 proto, const struct sk_buff *skb)
397{
398 struct nf_conntrack_tuple tuple;
399
Eric W. Biedermana31f1ad2015-09-18 14:33:04 -0500400 if (!nf_ct_get_tuplepr(skb, skb_network_offset(skb), proto, net, &tuple))
Joe Stringer7f8a4362015-08-26 11:31:48 -0700401 return NULL;
402 return __nf_ct_expect_find(net, zone, &tuple);
403}
404
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800405/* This replicates logic from nf_conntrack_core.c that is not exported. */
406static enum ip_conntrack_info
407ovs_ct_get_info(const struct nf_conntrack_tuple_hash *h)
408{
409 const struct nf_conn *ct = nf_ct_tuplehash_to_ctrack(h);
410
411 if (NF_CT_DIRECTION(h) == IP_CT_DIR_REPLY)
412 return IP_CT_ESTABLISHED_REPLY;
413 /* Once we've had two way comms, always ESTABLISHED. */
414 if (test_bit(IPS_SEEN_REPLY_BIT, &ct->status))
415 return IP_CT_ESTABLISHED;
416 if (test_bit(IPS_EXPECTED_BIT, &ct->status))
417 return IP_CT_RELATED;
418 return IP_CT_NEW;
419}
420
421/* Find an existing connection which this packet belongs to without
422 * re-attributing statistics or modifying the connection state. This allows an
423 * skb->nfct lost due to an upcall to be recovered during actions execution.
424 *
425 * Must be called with rcu_read_lock.
426 *
427 * On success, populates skb->nfct and skb->nfctinfo, and returns the
428 * connection. Returns NULL if there is no existing entry.
429 */
430static struct nf_conn *
431ovs_ct_find_existing(struct net *net, const struct nf_conntrack_zone *zone,
432 u8 l3num, struct sk_buff *skb)
433{
434 struct nf_conntrack_l3proto *l3proto;
435 struct nf_conntrack_l4proto *l4proto;
436 struct nf_conntrack_tuple tuple;
437 struct nf_conntrack_tuple_hash *h;
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800438 struct nf_conn *ct;
439 unsigned int dataoff;
440 u8 protonum;
441
442 l3proto = __nf_ct_l3proto_find(l3num);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800443 if (l3proto->get_l4proto(skb, skb_network_offset(skb), &dataoff,
444 &protonum) <= 0) {
445 pr_debug("ovs_ct_find_existing: Can't get protonum\n");
446 return NULL;
447 }
448 l4proto = __nf_ct_l4proto_find(l3num, protonum);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800449 if (!nf_ct_get_tuple(skb, skb_network_offset(skb), dataoff, l3num,
450 protonum, net, &tuple, l3proto, l4proto)) {
451 pr_debug("ovs_ct_find_existing: Can't get tuple\n");
452 return NULL;
453 }
454
455 /* look for tuple match */
456 h = nf_conntrack_find_get(net, zone, &tuple);
457 if (!h)
458 return NULL; /* Not found. */
459
460 ct = nf_ct_tuplehash_to_ctrack(h);
461
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800462 skb->nfct = &ct->ct_general;
Jarno Rajahalmebce91f82016-08-01 19:36:07 -0700463 skb->nfctinfo = ovs_ct_get_info(h);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800464 return ct;
465}
466
Joe Stringer7f8a4362015-08-26 11:31:48 -0700467/* Determine whether skb->nfct is equal to the result of conntrack lookup. */
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800468static bool skb_nfct_cached(struct net *net,
469 const struct sw_flow_key *key,
470 const struct ovs_conntrack_info *info,
471 struct sk_buff *skb)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700472{
473 enum ip_conntrack_info ctinfo;
474 struct nf_conn *ct;
475
476 ct = nf_ct_get(skb, &ctinfo);
Jarno Rajahalme289f2252016-03-10 10:54:20 -0800477 /* If no ct, check if we have evidence that an existing conntrack entry
478 * might be found for this skb. This happens when we lose a skb->nfct
479 * due to an upcall. If the connection was not confirmed, it is not
480 * cached and needs to be run through conntrack again.
481 */
482 if (!ct && key->ct.state & OVS_CS_F_TRACKED &&
483 !(key->ct.state & OVS_CS_F_INVALID) &&
484 key->ct.zone == info->zone.id)
485 ct = ovs_ct_find_existing(net, &info->zone, info->family, skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700486 if (!ct)
487 return false;
488 if (!net_eq(net, read_pnet(&ct->ct_net)))
489 return false;
490 if (!nf_ct_zone_equal_any(info->ct, nf_ct_zone(ct)))
491 return false;
Joe Stringercae3a262015-08-26 11:31:53 -0700492 if (info->helper) {
493 struct nf_conn_help *help;
494
495 help = nf_ct_ext_find(ct, NF_CT_EXT_HELPER);
496 if (help && rcu_access_pointer(help->helper) != info->helper)
497 return false;
498 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700499
500 return true;
501}
502
Jarno Rajahalme05752522016-03-10 10:54:23 -0800503#ifdef CONFIG_NF_NAT_NEEDED
504/* Modelled after nf_nat_ipv[46]_fn().
505 * range is only used for new, uninitialized NAT state.
506 * Returns either NF_ACCEPT or NF_DROP.
507 */
508static int ovs_ct_nat_execute(struct sk_buff *skb, struct nf_conn *ct,
509 enum ip_conntrack_info ctinfo,
510 const struct nf_nat_range *range,
511 enum nf_nat_manip_type maniptype)
512{
513 int hooknum, nh_off, err = NF_ACCEPT;
514
515 nh_off = skb_network_offset(skb);
Lance Richardson18767ac2017-01-12 19:33:18 -0500516 skb_pull_rcsum(skb, nh_off);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800517
518 /* See HOOK2MANIP(). */
519 if (maniptype == NF_NAT_MANIP_SRC)
520 hooknum = NF_INET_LOCAL_IN; /* Source NAT */
521 else
522 hooknum = NF_INET_LOCAL_OUT; /* Destination NAT */
523
524 switch (ctinfo) {
525 case IP_CT_RELATED:
526 case IP_CT_RELATED_REPLY:
Arnd Bergmann99b72482016-03-18 14:33:45 +0100527 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
528 skb->protocol == htons(ETH_P_IP) &&
Jarno Rajahalme05752522016-03-10 10:54:23 -0800529 ip_hdr(skb)->protocol == IPPROTO_ICMP) {
530 if (!nf_nat_icmp_reply_translation(skb, ct, ctinfo,
531 hooknum))
532 err = NF_DROP;
533 goto push;
Arnd Bergmann99b72482016-03-18 14:33:45 +0100534 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
535 skb->protocol == htons(ETH_P_IPV6)) {
Jarno Rajahalme05752522016-03-10 10:54:23 -0800536 __be16 frag_off;
537 u8 nexthdr = ipv6_hdr(skb)->nexthdr;
538 int hdrlen = ipv6_skip_exthdr(skb,
539 sizeof(struct ipv6hdr),
540 &nexthdr, &frag_off);
541
542 if (hdrlen >= 0 && nexthdr == IPPROTO_ICMPV6) {
543 if (!nf_nat_icmpv6_reply_translation(skb, ct,
544 ctinfo,
545 hooknum,
546 hdrlen))
547 err = NF_DROP;
548 goto push;
549 }
Jarno Rajahalme05752522016-03-10 10:54:23 -0800550 }
551 /* Non-ICMP, fall thru to initialize if needed. */
552 case IP_CT_NEW:
553 /* Seen it before? This can happen for loopback, retrans,
554 * or local packets.
555 */
556 if (!nf_nat_initialized(ct, maniptype)) {
557 /* Initialize according to the NAT action. */
558 err = (range && range->flags & NF_NAT_RANGE_MAP_IPS)
559 /* Action is set up to establish a new
560 * mapping.
561 */
562 ? nf_nat_setup_info(ct, range, maniptype)
563 : nf_nat_alloc_null_binding(ct, hooknum);
564 if (err != NF_ACCEPT)
565 goto push;
566 }
567 break;
568
569 case IP_CT_ESTABLISHED:
570 case IP_CT_ESTABLISHED_REPLY:
571 break;
572
573 default:
574 err = NF_DROP;
575 goto push;
576 }
577
578 err = nf_nat_packet(ct, ctinfo, hooknum, skb);
579push:
580 skb_push(skb, nh_off);
Lance Richardson18767ac2017-01-12 19:33:18 -0500581 skb_postpush_rcsum(skb, skb->data, nh_off);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800582
583 return err;
584}
585
586static void ovs_nat_update_key(struct sw_flow_key *key,
587 const struct sk_buff *skb,
588 enum nf_nat_manip_type maniptype)
589{
590 if (maniptype == NF_NAT_MANIP_SRC) {
591 __be16 src;
592
593 key->ct.state |= OVS_CS_F_SRC_NAT;
594 if (key->eth.type == htons(ETH_P_IP))
595 key->ipv4.addr.src = ip_hdr(skb)->saddr;
596 else if (key->eth.type == htons(ETH_P_IPV6))
597 memcpy(&key->ipv6.addr.src, &ipv6_hdr(skb)->saddr,
598 sizeof(key->ipv6.addr.src));
599 else
600 return;
601
602 if (key->ip.proto == IPPROTO_UDP)
603 src = udp_hdr(skb)->source;
604 else if (key->ip.proto == IPPROTO_TCP)
605 src = tcp_hdr(skb)->source;
606 else if (key->ip.proto == IPPROTO_SCTP)
607 src = sctp_hdr(skb)->source;
608 else
609 return;
610
611 key->tp.src = src;
612 } else {
613 __be16 dst;
614
615 key->ct.state |= OVS_CS_F_DST_NAT;
616 if (key->eth.type == htons(ETH_P_IP))
617 key->ipv4.addr.dst = ip_hdr(skb)->daddr;
618 else if (key->eth.type == htons(ETH_P_IPV6))
619 memcpy(&key->ipv6.addr.dst, &ipv6_hdr(skb)->daddr,
620 sizeof(key->ipv6.addr.dst));
621 else
622 return;
623
624 if (key->ip.proto == IPPROTO_UDP)
625 dst = udp_hdr(skb)->dest;
626 else if (key->ip.proto == IPPROTO_TCP)
627 dst = tcp_hdr(skb)->dest;
628 else if (key->ip.proto == IPPROTO_SCTP)
629 dst = sctp_hdr(skb)->dest;
630 else
631 return;
632
633 key->tp.dst = dst;
634 }
635}
636
637/* Returns NF_DROP if the packet should be dropped, NF_ACCEPT otherwise. */
638static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
639 const struct ovs_conntrack_info *info,
640 struct sk_buff *skb, struct nf_conn *ct,
641 enum ip_conntrack_info ctinfo)
642{
643 enum nf_nat_manip_type maniptype;
644 int err;
645
646 if (nf_ct_is_untracked(ct)) {
647 /* A NAT action may only be performed on tracked packets. */
648 return NF_ACCEPT;
649 }
650
651 /* Add NAT extension if not confirmed yet. */
652 if (!nf_ct_is_confirmed(ct) && !nf_ct_nat_ext_add(ct))
653 return NF_ACCEPT; /* Can't NAT. */
654
655 /* Determine NAT type.
656 * Check if the NAT type can be deduced from the tracked connection.
Jarno Rajahalme5745b0b2016-03-21 11:15:19 -0700657 * Make sure new expected connections (IP_CT_RELATED) are NATted only
658 * when committing.
Jarno Rajahalme05752522016-03-10 10:54:23 -0800659 */
660 if (info->nat & OVS_CT_NAT && ctinfo != IP_CT_NEW &&
661 ct->status & IPS_NAT_MASK &&
Jarno Rajahalme5745b0b2016-03-21 11:15:19 -0700662 (ctinfo != IP_CT_RELATED || info->commit)) {
Jarno Rajahalme05752522016-03-10 10:54:23 -0800663 /* NAT an established or related connection like before. */
664 if (CTINFO2DIR(ctinfo) == IP_CT_DIR_REPLY)
665 /* This is the REPLY direction for a connection
666 * for which NAT was applied in the forward
667 * direction. Do the reverse NAT.
668 */
669 maniptype = ct->status & IPS_SRC_NAT
670 ? NF_NAT_MANIP_DST : NF_NAT_MANIP_SRC;
671 else
672 maniptype = ct->status & IPS_SRC_NAT
673 ? NF_NAT_MANIP_SRC : NF_NAT_MANIP_DST;
674 } else if (info->nat & OVS_CT_SRC_NAT) {
675 maniptype = NF_NAT_MANIP_SRC;
676 } else if (info->nat & OVS_CT_DST_NAT) {
677 maniptype = NF_NAT_MANIP_DST;
678 } else {
679 return NF_ACCEPT; /* Connection is not NATed. */
680 }
681 err = ovs_ct_nat_execute(skb, ct, ctinfo, &info->range, maniptype);
682
683 /* Mark NAT done if successful and update the flow key. */
684 if (err == NF_ACCEPT)
685 ovs_nat_update_key(key, skb, maniptype);
686
687 return err;
688}
689#else /* !CONFIG_NF_NAT_NEEDED */
690static int ovs_ct_nat(struct net *net, struct sw_flow_key *key,
691 const struct ovs_conntrack_info *info,
692 struct sk_buff *skb, struct nf_conn *ct,
693 enum ip_conntrack_info ctinfo)
694{
695 return NF_ACCEPT;
696}
697#endif
698
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800699/* Pass 'skb' through conntrack in 'net', using zone configured in 'info', if
Jarno Rajahalme394e9102016-03-10 10:54:19 -0800700 * not done already. Update key with new CT state after passing the packet
701 * through conntrack.
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800702 * Note that if the packet is deemed invalid by conntrack, skb->nfct will be
703 * set to NULL and 0 will be returned.
704 */
Joe Stringer4f0909e2015-10-19 19:18:59 -0700705static int __ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
Joe Stringer7f8a4362015-08-26 11:31:48 -0700706 const struct ovs_conntrack_info *info,
707 struct sk_buff *skb)
708{
709 /* If we are recirculating packets to match on conntrack fields and
710 * committing with a separate conntrack action, then we don't need to
711 * actually run the packet through conntrack twice unless it's for a
712 * different zone.
713 */
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800714 bool cached = skb_nfct_cached(net, key, info, skb);
715 enum ip_conntrack_info ctinfo;
716 struct nf_conn *ct;
717
718 if (!cached) {
Joe Stringer7f8a4362015-08-26 11:31:48 -0700719 struct nf_conn *tmpl = info->ct;
Jarno Rajahalme5b6b9292016-03-10 10:54:21 -0800720 int err;
Joe Stringer7f8a4362015-08-26 11:31:48 -0700721
722 /* Associate skb with specified zone. */
723 if (tmpl) {
724 if (skb->nfct)
725 nf_conntrack_put(skb->nfct);
726 nf_conntrack_get(&tmpl->ct_general);
727 skb->nfct = &tmpl->ct_general;
728 skb->nfctinfo = IP_CT_NEW;
729 }
730
Jarno Rajahalme5b6b9292016-03-10 10:54:21 -0800731 /* Repeat if requested, see nf_iterate(). */
732 do {
733 err = nf_conntrack_in(net, info->family,
734 NF_INET_PRE_ROUTING, skb);
735 } while (err == NF_REPEAT);
736
737 if (err != NF_ACCEPT)
Joe Stringer7f8a4362015-08-26 11:31:48 -0700738 return -ENOENT;
Joe Stringercae3a262015-08-26 11:31:53 -0700739
Jarno Rajahalme05752522016-03-10 10:54:23 -0800740 /* Clear CT state NAT flags to mark that we have not yet done
741 * NAT after the nf_conntrack_in() call. We can actually clear
742 * the whole state, as it will be re-initialized below.
743 */
744 key->ct.state = 0;
745
746 /* Update the key, but keep the NAT flags. */
747 ovs_ct_update_key(skb, info, key, true, true);
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800748 }
Jarno Rajahalme394e9102016-03-10 10:54:19 -0800749
Jarno Rajahalme28b6e0c2016-03-10 10:54:22 -0800750 ct = nf_ct_get(skb, &ctinfo);
Jarno Rajahalme05752522016-03-10 10:54:23 -0800751 if (ct) {
752 /* Packets starting a new connection must be NATted before the
753 * helper, so that the helper knows about the NAT. We enforce
754 * this by delaying both NAT and helper calls for unconfirmed
755 * connections until the committing CT action. For later
756 * packets NAT and Helper may be called in either order.
757 *
758 * NAT will be done only if the CT action has NAT, and only
759 * once per packet (per zone), as guarded by the NAT bits in
760 * the key->ct.state.
761 */
762 if (info->nat && !(key->ct.state & OVS_CS_F_NAT_MASK) &&
763 (nf_ct_is_confirmed(ct) || info->commit) &&
764 ovs_ct_nat(net, key, info, skb, ct, ctinfo) != NF_ACCEPT) {
765 return -EINVAL;
766 }
767
Joe Stringer16ec3d42016-05-11 10:29:26 -0700768 /* Userspace may decide to perform a ct lookup without a helper
769 * specified followed by a (recirculate and) commit with one.
770 * Therefore, for unconfirmed connections which we will commit,
771 * we need to attach the helper here.
772 */
773 if (!nf_ct_is_confirmed(ct) && info->commit &&
774 info->helper && !nfct_help(ct)) {
775 int err = __nf_ct_try_assign_helper(ct, info->ct,
776 GFP_ATOMIC);
777 if (err)
778 return err;
779 }
780
Jarno Rajahalme05752522016-03-10 10:54:23 -0800781 /* Call the helper only if:
782 * - nf_conntrack_in() was executed above ("!cached") for a
783 * confirmed connection, or
784 * - When committing an unconfirmed connection.
785 */
786 if ((nf_ct_is_confirmed(ct) ? !cached : info->commit) &&
787 ovs_ct_helper(skb, info->family) != NF_ACCEPT) {
788 return -EINVAL;
789 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700790 }
791
792 return 0;
793}
794
795/* Lookup connection and read fields into key. */
796static int ovs_ct_lookup(struct net *net, struct sw_flow_key *key,
797 const struct ovs_conntrack_info *info,
798 struct sk_buff *skb)
799{
800 struct nf_conntrack_expect *exp;
801
Jarno Rajahalme9f13ded2016-03-10 10:54:18 -0800802 /* If we pass an expected packet through nf_conntrack_in() the
803 * expectation is typically removed, but the packet could still be
804 * lost in upcall processing. To prevent this from happening we
805 * perform an explicit expectation lookup. Expected connections are
806 * always new, and will be passed through conntrack only when they are
807 * committed, as it is OK to remove the expectation at that time.
808 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700809 exp = ovs_ct_expect_find(net, &info->zone, info->family, skb);
810 if (exp) {
811 u8 state;
812
Jarno Rajahalme05752522016-03-10 10:54:23 -0800813 /* NOTE: New connections are NATted and Helped only when
814 * committed, so we are not calling into NAT here.
815 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700816 state = OVS_CS_F_TRACKED | OVS_CS_F_NEW | OVS_CS_F_RELATED;
Joe Stringer182e3042015-08-26 11:31:49 -0700817 __ovs_ct_update_key(key, state, &info->zone, exp->master);
Samuel Gauthierd913d3a2016-06-28 17:22:26 +0200818 } else {
819 struct nf_conn *ct;
820 int err;
821
822 err = __ovs_ct_lookup(net, key, info, skb);
823 if (err)
824 return err;
825
826 ct = (struct nf_conn *)skb->nfct;
827 if (ct)
828 nf_ct_deliver_cached_events(ct);
829 }
Joe Stringer7f8a4362015-08-26 11:31:48 -0700830
831 return 0;
832}
833
Joe Stringer33db4122015-10-01 15:00:37 -0700834static bool labels_nonzero(const struct ovs_key_ct_labels *labels)
Joe Stringerc2ac6672015-08-26 11:31:52 -0700835{
836 size_t i;
837
Joe Stringer33db4122015-10-01 15:00:37 -0700838 for (i = 0; i < sizeof(*labels); i++)
839 if (labels->ct_labels[i])
Joe Stringerc2ac6672015-08-26 11:31:52 -0700840 return true;
841
842 return false;
843}
844
Jarno Rajahalme7d904c72016-06-21 14:59:38 -0700845/* Lookup connection and confirm if unconfirmed. */
846static int ovs_ct_commit(struct net *net, struct sw_flow_key *key,
847 const struct ovs_conntrack_info *info,
848 struct sk_buff *skb)
849{
850 int err;
851
852 err = __ovs_ct_lookup(net, key, info, skb);
853 if (err)
854 return err;
855
856 /* Apply changes before confirming the connection so that the initial
857 * conntrack NEW netlink event carries the values given in the CT
858 * action.
859 */
860 if (info->mark.mask) {
861 err = ovs_ct_set_mark(skb, key, info->mark.value,
862 info->mark.mask);
863 if (err)
864 return err;
865 }
866 if (labels_nonzero(&info->labels.mask)) {
867 err = ovs_ct_set_labels(skb, key, &info->labels.value,
868 &info->labels.mask);
869 if (err)
870 return err;
871 }
872 /* This will take care of sending queued events even if the connection
873 * is already confirmed.
874 */
875 if (nf_conntrack_confirm(skb) != NF_ACCEPT)
876 return -EINVAL;
877
878 return 0;
879}
880
Joe Stringer74c16612015-10-25 20:21:48 -0700881/* Returns 0 on success, -EINPROGRESS if 'skb' is stolen, or other nonzero
882 * value if 'skb' is freed.
883 */
Joe Stringer7f8a4362015-08-26 11:31:48 -0700884int ovs_ct_execute(struct net *net, struct sk_buff *skb,
885 struct sw_flow_key *key,
886 const struct ovs_conntrack_info *info)
887{
888 int nh_ofs;
889 int err;
890
891 /* The conntrack module expects to be working at L3. */
892 nh_ofs = skb_network_offset(skb);
Lance Richardson18767ac2017-01-12 19:33:18 -0500893 skb_pull_rcsum(skb, nh_ofs);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700894
895 if (key->ip.frag != OVS_FRAG_TYPE_NONE) {
896 err = handle_fragments(net, key, info->zone.id, skb);
897 if (err)
898 return err;
899 }
900
Joe Stringerab38a7b2015-10-06 11:00:01 -0700901 if (info->commit)
Jarno Rajahalme7d904c72016-06-21 14:59:38 -0700902 err = ovs_ct_commit(net, key, info, skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700903 else
904 err = ovs_ct_lookup(net, key, info, skb);
905
906 skb_push(skb, nh_ofs);
Lance Richardson18767ac2017-01-12 19:33:18 -0500907 skb_postpush_rcsum(skb, skb->data, nh_ofs);
Joe Stringer74c16612015-10-25 20:21:48 -0700908 if (err)
909 kfree_skb(skb);
Joe Stringer7f8a4362015-08-26 11:31:48 -0700910 return err;
911}
912
Joe Stringercae3a262015-08-26 11:31:53 -0700913static int ovs_ct_add_helper(struct ovs_conntrack_info *info, const char *name,
914 const struct sw_flow_key *key, bool log)
915{
916 struct nf_conntrack_helper *helper;
917 struct nf_conn_help *help;
918
919 helper = nf_conntrack_helper_try_module_get(name, info->family,
920 key->ip.proto);
921 if (!helper) {
922 OVS_NLERR(log, "Unknown helper \"%s\"", name);
923 return -EINVAL;
924 }
925
926 help = nf_ct_helper_ext_add(info->ct, helper, GFP_KERNEL);
927 if (!help) {
928 module_put(helper->me);
929 return -ENOMEM;
930 }
931
932 rcu_assign_pointer(help->helper, helper);
933 info->helper = helper;
934 return 0;
935}
936
Jarno Rajahalme05752522016-03-10 10:54:23 -0800937#ifdef CONFIG_NF_NAT_NEEDED
938static int parse_nat(const struct nlattr *attr,
939 struct ovs_conntrack_info *info, bool log)
940{
941 struct nlattr *a;
942 int rem;
943 bool have_ip_max = false;
944 bool have_proto_max = false;
945 bool ip_vers = (info->family == NFPROTO_IPV6);
946
947 nla_for_each_nested(a, attr, rem) {
948 static const int ovs_nat_attr_lens[OVS_NAT_ATTR_MAX + 1][2] = {
949 [OVS_NAT_ATTR_SRC] = {0, 0},
950 [OVS_NAT_ATTR_DST] = {0, 0},
951 [OVS_NAT_ATTR_IP_MIN] = {sizeof(struct in_addr),
952 sizeof(struct in6_addr)},
953 [OVS_NAT_ATTR_IP_MAX] = {sizeof(struct in_addr),
954 sizeof(struct in6_addr)},
955 [OVS_NAT_ATTR_PROTO_MIN] = {sizeof(u16), sizeof(u16)},
956 [OVS_NAT_ATTR_PROTO_MAX] = {sizeof(u16), sizeof(u16)},
957 [OVS_NAT_ATTR_PERSISTENT] = {0, 0},
958 [OVS_NAT_ATTR_PROTO_HASH] = {0, 0},
959 [OVS_NAT_ATTR_PROTO_RANDOM] = {0, 0},
960 };
961 int type = nla_type(a);
962
963 if (type > OVS_NAT_ATTR_MAX) {
964 OVS_NLERR(log,
965 "Unknown NAT attribute (type=%d, max=%d).\n",
966 type, OVS_NAT_ATTR_MAX);
967 return -EINVAL;
968 }
969
970 if (nla_len(a) != ovs_nat_attr_lens[type][ip_vers]) {
971 OVS_NLERR(log,
972 "NAT attribute type %d has unexpected length (%d != %d).\n",
973 type, nla_len(a),
974 ovs_nat_attr_lens[type][ip_vers]);
975 return -EINVAL;
976 }
977
978 switch (type) {
979 case OVS_NAT_ATTR_SRC:
980 case OVS_NAT_ATTR_DST:
981 if (info->nat) {
982 OVS_NLERR(log,
983 "Only one type of NAT may be specified.\n"
984 );
985 return -ERANGE;
986 }
987 info->nat |= OVS_CT_NAT;
988 info->nat |= ((type == OVS_NAT_ATTR_SRC)
989 ? OVS_CT_SRC_NAT : OVS_CT_DST_NAT);
990 break;
991
992 case OVS_NAT_ATTR_IP_MIN:
Haishuang Yanac71b462016-03-28 18:08:59 +0800993 nla_memcpy(&info->range.min_addr, a,
994 sizeof(info->range.min_addr));
Jarno Rajahalme05752522016-03-10 10:54:23 -0800995 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
996 break;
997
998 case OVS_NAT_ATTR_IP_MAX:
999 have_ip_max = true;
1000 nla_memcpy(&info->range.max_addr, a,
1001 sizeof(info->range.max_addr));
1002 info->range.flags |= NF_NAT_RANGE_MAP_IPS;
1003 break;
1004
1005 case OVS_NAT_ATTR_PROTO_MIN:
1006 info->range.min_proto.all = htons(nla_get_u16(a));
1007 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1008 break;
1009
1010 case OVS_NAT_ATTR_PROTO_MAX:
1011 have_proto_max = true;
1012 info->range.max_proto.all = htons(nla_get_u16(a));
1013 info->range.flags |= NF_NAT_RANGE_PROTO_SPECIFIED;
1014 break;
1015
1016 case OVS_NAT_ATTR_PERSISTENT:
1017 info->range.flags |= NF_NAT_RANGE_PERSISTENT;
1018 break;
1019
1020 case OVS_NAT_ATTR_PROTO_HASH:
1021 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM;
1022 break;
1023
1024 case OVS_NAT_ATTR_PROTO_RANDOM:
1025 info->range.flags |= NF_NAT_RANGE_PROTO_RANDOM_FULLY;
1026 break;
1027
1028 default:
1029 OVS_NLERR(log, "Unknown nat attribute (%d).\n", type);
1030 return -EINVAL;
1031 }
1032 }
1033
1034 if (rem > 0) {
1035 OVS_NLERR(log, "NAT attribute has %d unknown bytes.\n", rem);
1036 return -EINVAL;
1037 }
1038 if (!info->nat) {
1039 /* Do not allow flags if no type is given. */
1040 if (info->range.flags) {
1041 OVS_NLERR(log,
1042 "NAT flags may be given only when NAT range (SRC or DST) is also specified.\n"
1043 );
1044 return -EINVAL;
1045 }
1046 info->nat = OVS_CT_NAT; /* NAT existing connections. */
1047 } else if (!info->commit) {
1048 OVS_NLERR(log,
1049 "NAT attributes may be specified only when CT COMMIT flag is also specified.\n"
1050 );
1051 return -EINVAL;
1052 }
1053 /* Allow missing IP_MAX. */
1054 if (info->range.flags & NF_NAT_RANGE_MAP_IPS && !have_ip_max) {
1055 memcpy(&info->range.max_addr, &info->range.min_addr,
1056 sizeof(info->range.max_addr));
1057 }
1058 /* Allow missing PROTO_MAX. */
1059 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1060 !have_proto_max) {
1061 info->range.max_proto.all = info->range.min_proto.all;
1062 }
1063 return 0;
1064}
1065#endif
1066
Joe Stringer7f8a4362015-08-26 11:31:48 -07001067static const struct ovs_ct_len_tbl ovs_ct_attr_lens[OVS_CT_ATTR_MAX + 1] = {
Joe Stringerab38a7b2015-10-06 11:00:01 -07001068 [OVS_CT_ATTR_COMMIT] = { .minlen = 0, .maxlen = 0 },
Joe Stringer7f8a4362015-08-26 11:31:48 -07001069 [OVS_CT_ATTR_ZONE] = { .minlen = sizeof(u16),
1070 .maxlen = sizeof(u16) },
Joe Stringer182e3042015-08-26 11:31:49 -07001071 [OVS_CT_ATTR_MARK] = { .minlen = sizeof(struct md_mark),
1072 .maxlen = sizeof(struct md_mark) },
Joe Stringer33db4122015-10-01 15:00:37 -07001073 [OVS_CT_ATTR_LABELS] = { .minlen = sizeof(struct md_labels),
1074 .maxlen = sizeof(struct md_labels) },
Joe Stringercae3a262015-08-26 11:31:53 -07001075 [OVS_CT_ATTR_HELPER] = { .minlen = 1,
Jarno Rajahalme05752522016-03-10 10:54:23 -08001076 .maxlen = NF_CT_HELPER_NAME_LEN },
1077#ifdef CONFIG_NF_NAT_NEEDED
1078 /* NAT length is checked when parsing the nested attributes. */
1079 [OVS_CT_ATTR_NAT] = { .minlen = 0, .maxlen = INT_MAX },
1080#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001081};
1082
1083static int parse_ct(const struct nlattr *attr, struct ovs_conntrack_info *info,
Joe Stringercae3a262015-08-26 11:31:53 -07001084 const char **helper, bool log)
Joe Stringer7f8a4362015-08-26 11:31:48 -07001085{
1086 struct nlattr *a;
1087 int rem;
1088
1089 nla_for_each_nested(a, attr, rem) {
1090 int type = nla_type(a);
1091 int maxlen = ovs_ct_attr_lens[type].maxlen;
1092 int minlen = ovs_ct_attr_lens[type].minlen;
1093
1094 if (type > OVS_CT_ATTR_MAX) {
1095 OVS_NLERR(log,
1096 "Unknown conntrack attr (type=%d, max=%d)",
1097 type, OVS_CT_ATTR_MAX);
1098 return -EINVAL;
1099 }
1100 if (nla_len(a) < minlen || nla_len(a) > maxlen) {
1101 OVS_NLERR(log,
1102 "Conntrack attr type has unexpected length (type=%d, length=%d, expected=%d)",
1103 type, nla_len(a), maxlen);
1104 return -EINVAL;
1105 }
1106
1107 switch (type) {
Joe Stringerab38a7b2015-10-06 11:00:01 -07001108 case OVS_CT_ATTR_COMMIT:
1109 info->commit = true;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001110 break;
1111#ifdef CONFIG_NF_CONNTRACK_ZONES
1112 case OVS_CT_ATTR_ZONE:
1113 info->zone.id = nla_get_u16(a);
1114 break;
1115#endif
Joe Stringer182e3042015-08-26 11:31:49 -07001116#ifdef CONFIG_NF_CONNTRACK_MARK
1117 case OVS_CT_ATTR_MARK: {
1118 struct md_mark *mark = nla_data(a);
1119
Joe Stringere754ec62015-10-19 19:19:00 -07001120 if (!mark->mask) {
1121 OVS_NLERR(log, "ct_mark mask cannot be 0");
1122 return -EINVAL;
1123 }
Joe Stringer182e3042015-08-26 11:31:49 -07001124 info->mark = *mark;
1125 break;
1126 }
1127#endif
Joe Stringerc2ac6672015-08-26 11:31:52 -07001128#ifdef CONFIG_NF_CONNTRACK_LABELS
Joe Stringer33db4122015-10-01 15:00:37 -07001129 case OVS_CT_ATTR_LABELS: {
1130 struct md_labels *labels = nla_data(a);
Joe Stringerc2ac6672015-08-26 11:31:52 -07001131
Joe Stringere754ec62015-10-19 19:19:00 -07001132 if (!labels_nonzero(&labels->mask)) {
1133 OVS_NLERR(log, "ct_labels mask cannot be 0");
1134 return -EINVAL;
1135 }
Joe Stringer33db4122015-10-01 15:00:37 -07001136 info->labels = *labels;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001137 break;
1138 }
1139#endif
Joe Stringercae3a262015-08-26 11:31:53 -07001140 case OVS_CT_ATTR_HELPER:
1141 *helper = nla_data(a);
1142 if (!memchr(*helper, '\0', nla_len(a))) {
1143 OVS_NLERR(log, "Invalid conntrack helper");
1144 return -EINVAL;
1145 }
1146 break;
Jarno Rajahalme05752522016-03-10 10:54:23 -08001147#ifdef CONFIG_NF_NAT_NEEDED
1148 case OVS_CT_ATTR_NAT: {
1149 int err = parse_nat(a, info, log);
1150
1151 if (err)
1152 return err;
1153 break;
1154 }
1155#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001156 default:
1157 OVS_NLERR(log, "Unknown conntrack attr (%d)",
1158 type);
1159 return -EINVAL;
1160 }
1161 }
1162
Jarno Rajahalme7d904c72016-06-21 14:59:38 -07001163#ifdef CONFIG_NF_CONNTRACK_MARK
1164 if (!info->commit && info->mark.mask) {
1165 OVS_NLERR(log,
1166 "Setting conntrack mark requires 'commit' flag.");
1167 return -EINVAL;
1168 }
1169#endif
1170#ifdef CONFIG_NF_CONNTRACK_LABELS
1171 if (!info->commit && labels_nonzero(&info->labels.mask)) {
1172 OVS_NLERR(log,
1173 "Setting conntrack labels requires 'commit' flag.");
1174 return -EINVAL;
1175 }
1176#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001177 if (rem > 0) {
1178 OVS_NLERR(log, "Conntrack attr has %d unknown bytes", rem);
1179 return -EINVAL;
1180 }
1181
1182 return 0;
1183}
1184
Joe Stringerc2ac6672015-08-26 11:31:52 -07001185bool ovs_ct_verify(struct net *net, enum ovs_key_attr attr)
Joe Stringer7f8a4362015-08-26 11:31:48 -07001186{
1187 if (attr == OVS_KEY_ATTR_CT_STATE)
1188 return true;
1189 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1190 attr == OVS_KEY_ATTR_CT_ZONE)
1191 return true;
Joe Stringer182e3042015-08-26 11:31:49 -07001192 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) &&
1193 attr == OVS_KEY_ATTR_CT_MARK)
1194 return true;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001195 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringer33db4122015-10-01 15:00:37 -07001196 attr == OVS_KEY_ATTR_CT_LABELS) {
Joe Stringerc2ac6672015-08-26 11:31:52 -07001197 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1198
1199 return ovs_net->xt_label;
1200 }
Joe Stringer7f8a4362015-08-26 11:31:48 -07001201
1202 return false;
1203}
1204
1205int ovs_ct_copy_action(struct net *net, const struct nlattr *attr,
1206 const struct sw_flow_key *key,
1207 struct sw_flow_actions **sfa, bool log)
1208{
1209 struct ovs_conntrack_info ct_info;
Joe Stringercae3a262015-08-26 11:31:53 -07001210 const char *helper = NULL;
Joe Stringer7f8a4362015-08-26 11:31:48 -07001211 u16 family;
1212 int err;
1213
1214 family = key_to_nfproto(key);
1215 if (family == NFPROTO_UNSPEC) {
1216 OVS_NLERR(log, "ct family unspecified");
1217 return -EINVAL;
1218 }
1219
1220 memset(&ct_info, 0, sizeof(ct_info));
1221 ct_info.family = family;
1222
1223 nf_ct_zone_init(&ct_info.zone, NF_CT_DEFAULT_ZONE_ID,
1224 NF_CT_DEFAULT_ZONE_DIR, 0);
1225
Joe Stringercae3a262015-08-26 11:31:53 -07001226 err = parse_ct(attr, &ct_info, &helper, log);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001227 if (err)
1228 return err;
1229
1230 /* Set up template for tracking connections in specific zones. */
1231 ct_info.ct = nf_ct_tmpl_alloc(net, &ct_info.zone, GFP_KERNEL);
1232 if (!ct_info.ct) {
1233 OVS_NLERR(log, "Failed to allocate conntrack template");
1234 return -ENOMEM;
1235 }
Joe Stringer90c7afc962015-12-23 14:39:27 -08001236
1237 __set_bit(IPS_CONFIRMED_BIT, &ct_info.ct->status);
1238 nf_conntrack_get(&ct_info.ct->ct_general);
1239
Joe Stringercae3a262015-08-26 11:31:53 -07001240 if (helper) {
1241 err = ovs_ct_add_helper(&ct_info, helper, key, log);
1242 if (err)
1243 goto err_free_ct;
1244 }
Joe Stringer7f8a4362015-08-26 11:31:48 -07001245
1246 err = ovs_nla_add_action(sfa, OVS_ACTION_ATTR_CT, &ct_info,
1247 sizeof(ct_info), log);
1248 if (err)
1249 goto err_free_ct;
1250
Joe Stringer7f8a4362015-08-26 11:31:48 -07001251 return 0;
1252err_free_ct:
Joe Stringer2f3ab9f2015-12-09 14:07:39 -08001253 __ovs_ct_free_action(&ct_info);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001254 return err;
1255}
1256
Jarno Rajahalme05752522016-03-10 10:54:23 -08001257#ifdef CONFIG_NF_NAT_NEEDED
1258static bool ovs_ct_nat_to_attr(const struct ovs_conntrack_info *info,
1259 struct sk_buff *skb)
1260{
1261 struct nlattr *start;
1262
1263 start = nla_nest_start(skb, OVS_CT_ATTR_NAT);
1264 if (!start)
1265 return false;
1266
1267 if (info->nat & OVS_CT_SRC_NAT) {
1268 if (nla_put_flag(skb, OVS_NAT_ATTR_SRC))
1269 return false;
1270 } else if (info->nat & OVS_CT_DST_NAT) {
1271 if (nla_put_flag(skb, OVS_NAT_ATTR_DST))
1272 return false;
1273 } else {
1274 goto out;
1275 }
1276
1277 if (info->range.flags & NF_NAT_RANGE_MAP_IPS) {
Arnd Bergmann99b72482016-03-18 14:33:45 +01001278 if (IS_ENABLED(CONFIG_NF_NAT_IPV4) &&
1279 info->family == NFPROTO_IPV4) {
Jarno Rajahalme05752522016-03-10 10:54:23 -08001280 if (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MIN,
1281 info->range.min_addr.ip) ||
1282 (info->range.max_addr.ip
1283 != info->range.min_addr.ip &&
1284 (nla_put_in_addr(skb, OVS_NAT_ATTR_IP_MAX,
1285 info->range.max_addr.ip))))
1286 return false;
Arnd Bergmann99b72482016-03-18 14:33:45 +01001287 } else if (IS_ENABLED(CONFIG_NF_NAT_IPV6) &&
1288 info->family == NFPROTO_IPV6) {
Jarno Rajahalme05752522016-03-10 10:54:23 -08001289 if (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MIN,
1290 &info->range.min_addr.in6) ||
1291 (memcmp(&info->range.max_addr.in6,
1292 &info->range.min_addr.in6,
1293 sizeof(info->range.max_addr.in6)) &&
1294 (nla_put_in6_addr(skb, OVS_NAT_ATTR_IP_MAX,
1295 &info->range.max_addr.in6))))
1296 return false;
Jarno Rajahalme05752522016-03-10 10:54:23 -08001297 } else {
1298 return false;
1299 }
1300 }
1301 if (info->range.flags & NF_NAT_RANGE_PROTO_SPECIFIED &&
1302 (nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MIN,
1303 ntohs(info->range.min_proto.all)) ||
1304 (info->range.max_proto.all != info->range.min_proto.all &&
1305 nla_put_u16(skb, OVS_NAT_ATTR_PROTO_MAX,
1306 ntohs(info->range.max_proto.all)))))
1307 return false;
1308
1309 if (info->range.flags & NF_NAT_RANGE_PERSISTENT &&
1310 nla_put_flag(skb, OVS_NAT_ATTR_PERSISTENT))
1311 return false;
1312 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM &&
1313 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_HASH))
1314 return false;
1315 if (info->range.flags & NF_NAT_RANGE_PROTO_RANDOM_FULLY &&
1316 nla_put_flag(skb, OVS_NAT_ATTR_PROTO_RANDOM))
1317 return false;
1318out:
1319 nla_nest_end(skb, start);
1320
1321 return true;
1322}
1323#endif
1324
Joe Stringer7f8a4362015-08-26 11:31:48 -07001325int ovs_ct_action_to_attr(const struct ovs_conntrack_info *ct_info,
1326 struct sk_buff *skb)
1327{
1328 struct nlattr *start;
1329
1330 start = nla_nest_start(skb, OVS_ACTION_ATTR_CT);
1331 if (!start)
1332 return -EMSGSIZE;
1333
Joe Stringerab38a7b2015-10-06 11:00:01 -07001334 if (ct_info->commit && nla_put_flag(skb, OVS_CT_ATTR_COMMIT))
Joe Stringer7f8a4362015-08-26 11:31:48 -07001335 return -EMSGSIZE;
1336 if (IS_ENABLED(CONFIG_NF_CONNTRACK_ZONES) &&
1337 nla_put_u16(skb, OVS_CT_ATTR_ZONE, ct_info->zone.id))
1338 return -EMSGSIZE;
Joe Stringere754ec62015-10-19 19:19:00 -07001339 if (IS_ENABLED(CONFIG_NF_CONNTRACK_MARK) && ct_info->mark.mask &&
Joe Stringer182e3042015-08-26 11:31:49 -07001340 nla_put(skb, OVS_CT_ATTR_MARK, sizeof(ct_info->mark),
1341 &ct_info->mark))
1342 return -EMSGSIZE;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001343 if (IS_ENABLED(CONFIG_NF_CONNTRACK_LABELS) &&
Joe Stringere754ec62015-10-19 19:19:00 -07001344 labels_nonzero(&ct_info->labels.mask) &&
Joe Stringer33db4122015-10-01 15:00:37 -07001345 nla_put(skb, OVS_CT_ATTR_LABELS, sizeof(ct_info->labels),
1346 &ct_info->labels))
Joe Stringerc2ac6672015-08-26 11:31:52 -07001347 return -EMSGSIZE;
Joe Stringercae3a262015-08-26 11:31:53 -07001348 if (ct_info->helper) {
1349 if (nla_put_string(skb, OVS_CT_ATTR_HELPER,
1350 ct_info->helper->name))
1351 return -EMSGSIZE;
1352 }
Jarno Rajahalme05752522016-03-10 10:54:23 -08001353#ifdef CONFIG_NF_NAT_NEEDED
1354 if (ct_info->nat && !ovs_ct_nat_to_attr(ct_info, skb))
1355 return -EMSGSIZE;
1356#endif
Joe Stringer7f8a4362015-08-26 11:31:48 -07001357 nla_nest_end(skb, start);
1358
1359 return 0;
1360}
1361
1362void ovs_ct_free_action(const struct nlattr *a)
1363{
1364 struct ovs_conntrack_info *ct_info = nla_data(a);
1365
Joe Stringer2f3ab9f2015-12-09 14:07:39 -08001366 __ovs_ct_free_action(ct_info);
1367}
1368
1369static void __ovs_ct_free_action(struct ovs_conntrack_info *ct_info)
1370{
Joe Stringercae3a262015-08-26 11:31:53 -07001371 if (ct_info->helper)
1372 module_put(ct_info->helper->me);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001373 if (ct_info->ct)
Joe Stringer76644232016-09-01 18:01:47 -07001374 nf_ct_tmpl_free(ct_info->ct);
Joe Stringer7f8a4362015-08-26 11:31:48 -07001375}
Joe Stringerc2ac6672015-08-26 11:31:52 -07001376
1377void ovs_ct_init(struct net *net)
1378{
Joe Stringer33db4122015-10-01 15:00:37 -07001379 unsigned int n_bits = sizeof(struct ovs_key_ct_labels) * BITS_PER_BYTE;
Joe Stringerc2ac6672015-08-26 11:31:52 -07001380 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1381
Florian Westphaladff6c62016-04-12 18:14:25 +02001382 if (nf_connlabels_get(net, n_bits - 1)) {
Joe Stringerc2ac6672015-08-26 11:31:52 -07001383 ovs_net->xt_label = false;
1384 OVS_NLERR(true, "Failed to set connlabel length");
1385 } else {
1386 ovs_net->xt_label = true;
1387 }
1388}
1389
1390void ovs_ct_exit(struct net *net)
1391{
1392 struct ovs_net *ovs_net = net_generic(net, ovs_net_id);
1393
1394 if (ovs_net->xt_label)
1395 nf_connlabels_put(net);
1396}